Looking for the H alpha Raman wings in SMC-N66 (NGC 346)

Start off the same as the other notebooks - import the libraries and load the data

Deal with the sky correction

This is a step that is particular to the NGC 346 dataset, and hopefully will not be necessary for other regions.

I cannot use the same procedure developed in the 03-00-ha-moment-maps notebook of subtracting off the spectrum from a region where the sky oversubtraction is most apparent. When I tried it, it just increased the noise too much in the wings.

What I will try instead is to simply mask out regions where the Ha line goes negative.

Inspect the data cube

The full data cube is in cube, while hacube is a 600 Å window around Hα.

Looking at the data cube in DS9 I found a region that looks promising fro the Raman wings, so we will look at that first:

Images of different wavelength ranges

So the red wing and blue wing look like they are definitely there. The red wing can be seen on both sides of the [N II] λ6583 line. In Orion, we only see it clearly for $\lambda > 6600$. Perhaps this is because the C II λ6578 line is weaker in the LMC.

First, make an image of the outer red wing:

We see filamentary structure that is similar to the other emission lines (see 11-line-profiles notebook). Note that I am masking out the stars by using a condition on cont6600. It is a delicate balancing act between cutting out the PSF wings of the srars without losing too much of the nebular emission.

The map is very noisy but we can try and improve things by rebinning to 8x8. I also take the opportunity to trim off a 10-pixel margin all around, since there are some very noisy pixels there.

This looks a lot better. We can see the central filament a lot more clearly.

Extract spectrum for rectangular regions

Make some box regions for extracting the spectra. I use the astropy affiliated regions package (see docs)

I make four boxes. The first two are in regions where the Raman wing is strong, while the second two are where it is weak.

Plot an image of the entire bandbass in pixel coordinates and plot the boxes on top of it:

We can get the pixel slices from each box like this:

So we can extract the spectrum for each box. We apply the wing mask that we used above to the cube. (This is a much simpler way of combining a 2D and 3D mask).

I plot the spectrum for each box and also fit the continuum, so we can easily see if there are any Raman wings present.

The two boxes that were selected to cover the central filament (blue and red) show clear wings. Better seen on the red side, but also there on the blue.

The other two boxes, which are off the filament (magenta and green) show no Raman wings at all.

The thick yellow lines show the wavelength sections that are used for fitting the continuum (3rd-order polynomial).

I have put vertical lines at the wavelengths of the 6633 and 6664 features, assuming redshift of 160 km/s. The 6633 feature is in the middle of two emission lines. What are they? Could be night sky airglow lines

Choose suitable bands to measure the Raman wings

In the Orion paper, we have the 3 closest bands in the red wing: R040, R058, R087. These are marked by pink boxes in the previous figure.

In order to stand a chance of getting good maps, we need to get a better estimate of the continuum than the cont6600 that we used above, since this includes part of the Raman wing that we want to measure.

On the other hand, fitting a 3rd of 4th order polynomial, like we just did for the rectangular boxes is not practical since the individual pixels are too noisy.

A compromise would be to fit a linear trend between the continuum around 6400 and the continuum around 6700.

The above figure shows that the continuum is approximately linear over this range. We should check if there are any absorption edges that might cause discontinuities in the continuum

That seems to have worked OK. Now we need to do this for every pixel.

Fit the continuum pixel-by-pixel to the whole cube

This is ridiculously slow. I need to rewrite it to do my own continuum fitting. It turns out that the problem is that iterating over the individual Spectrum objects in a Cube is very inefficient. So I have to make sure to iterate over just the .data components, which are numpy arrays.

Copy the limits of the Raman bands from the Orion project:

Continuum-subtracted images of the red bands

So, we can se something in all the bands, but the first two are best.

I am in two minds whether to mask out the point sources or not

Continuum-subtracted images of the blue bands

The last band does not show anything, but the other three do.

Ratio of wing to core

This shows a clear difference in distribution between the wings and the core.

Look at the very brightest Raman pixels

These are all close to stars, but they are not actually stellar emission.

We see the Raman absorption lines!!

Note that this spectrum is dominated by the bright knot seen in the emission line maps, which seems to be associated with a young high-mass star.

There are two other bright knots too.

The total brightness of the Raman wings from these knots is higher than that of the diffuse emission from the entire rest of the nebula (see below).

Note that multiplying a Cube by a constant destroys the mask, so we need to do the sum before multiplying

Another thing – some of this flux comes from some very bright pixels towards the edge of the map. I had originally masked them all out (see the comment # Trim off noisy bands close to edges above), but I have reinstated the bottom and left edges, since they made quite a difference.

Look at the diffuse but still bright Raman pixels

Sources of interest

  1. The bright knot (j, i, = 147, 122)
  2. The cometary globule (j, i = 180, 109)
  3. The corner globule (j, i = 40, 40) – has a faint embedded Hα point source with high linewidth, but the Ramn scattering is more extended.
  4. The blue ball (j, i = 99, 59). Much brighter in the innermost bands. Perhaps it is not Raman scattering at all, but is a fast outflow?

Actually, I will come back to this later. It would be easier to take notes in the org file.